[389-users] changelog deadlock replication failures with DNA

[Mahadevan, Venkat]

> If the operation fails to write into the changelog, the operation fails. In your case, it means that the ldapclient should receive an error.
> So it is like if the operation never happened and is not replicated.

Hi Thierry,

That’s what I would expect to, but that does not seem to be the case. In my access log on the master server below,
the client successfully adds the entry to the master server and receives an error code return of 0:

[12/Jun/2013:13:36:59 -0700] conn=19048 op=0 BIND dn="cn=Directory Manager" method=128 version=3
[12/Jun/2013:13:36:59 -0700] conn=19048 op=0 RESULT err=0 tag=97 nentries=0 etime=0 dn="cn=directory manager"
[12/Jun/2013:13:36:59 -0700] conn=19048 op=1 ADD dn="uid=jmeter325,dc=tst,dc=id,dc=ubc,dc=ca"
[12/Jun/2013:13:37:04 -0700] conn=19048 op=1 RESULT err=0 tag=105 nentries=0 etime=5 csn=51b8dbec002b02bd0000

In the error log at the same time the above happens:

[12/Jun/2013:13:37:02 -0700] NSMMReplicationPlugin - replica_replace_ruv_tombstone: failed to update replication update vector for replica dc=tst,dc=id,dc=ubc,dc=ca: LDAP error - 51
[12/Jun/2013:13:37:04 -0700] NSMMReplicationPlugin - changelog program - _cl5WriteOperationTxn: retry (49) the transaction (csn=51b8dbec002b02bd0000) failed (rc=-30994 (DB_LOCK_DEADLOCK: Locker killed to resolve a deadlock))
[12/Jun/2013:13:37:04 -0700] NSMMReplicationPlugin - changelog program - _cl5WriteOperationTxn: failed to write entry with csn (51b8dbec002b02bd0000); db error - -30994 DB_LOCK_DEADLOCK: Locker killed to resolve a deadlock
[12/Jun/2013:13:37:04 -0700] NSMMReplicationPlugin - write_changelog_and_ruv: can't add a change for uid=jmeter325,dc=tst,dc=id,dc=ubc,dc=ca (uniqid: c0a7827c-d39f11e2-879add5e-44f44922, optype: 16) to changelog csn 51b8dbec002b02bd0000

The net result is that the entry is successfully added to the master server but will never replication to any of the consumers.

cheers,

VM
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.fedoraproject.org/pipermail/389-users/attachments/20130614/0d0145f8/attachment.html>